Skip to main content
LLM Security Database
Skip to research search
Last analyzed Sep 9, 2026

Language Model Security Database

985 research findings · 1119 evaluated models

Filtered research findings

468 entries

Matches every word across titles, descriptions, sources, affected systems, and models.

Self-Jailbreak Safety Override

Published Analyzed Source: arXiv

Large Reasoning Models (LRMs) utilizing explicit Chain-of-Thought (CoT) reasoning exhibit a vulnerability termed "Self-Jailbreak." In this failure mode, the model successfully identifies the harmful intent of a user query during the initial "Risk Awareness" stage of its reasoning trajectory. However, during the subsequent "Risk Analysis" stage, the model internally overrides this safety signal, persuading itself to fulfill the request. This override typically occurs through cognitive patterns…

When Models Outthink Their Safety: Unveiling and Mitigating Self-Jailbreak in Large Reasoning Models
Evaluated models: o1, DeepSeek R1, DeepSeek R1 0528+3 more
  • o1
  • DeepSeek R1
  • DeepSeek R1 0528
  • Qwen 3 8B
  • Qwen 3 14B
  • Qwen 3 32B

Benign Reasoning Self-Jailbreaks LLMs

Published Analyzed Source: arXiv

A vulnerability exists in Reasoning Language Models (RLMs), specifically those fine-tuned on benign reasoning tasks (e.g., mathematics, coding) such as DeepSeek-R1-distilled, s1.1, and Phi-4-mini-reasoning. The vulnerability, termed "Self-Jailbreaking," allows the model to circumvent its own safety guardrails during the internal Chain-of-Thought (CoT) generation process. Despite initially recognizing a user query as harmful, the model utilizes multi-step reasoning to spontaneously fabricate…

Self-Jailbreaking: Language Models Can Reason Themselves Out of Safety Alignment After Benign Reasoning Training
Evaluated models: s1.1 7B, Qwen 2.5 7B Instruct, DeepSeek R1 Distill Qwen 1.5B+6 more
  • s1.1 7B
  • Qwen 2.5 7B Instruct
  • DeepSeek R1 Distill Qwen 1.5B
  • DeepSeek R1 Distill Qwen 7B
  • DeepSeek R1 Distill Qwen 14B
  • DeepSeek R1 Distill Qwen 32B
  • DeepSeek R1 Distill Llama 8B
  • Phi-4 Mini Reasoning
  • Nemotron Research Reasoning

Trust Paradox Agent Over-Exposure

Published Analyzed Source: arXiv

A privilege escalation and information disclosure vulnerability exists in Large Language Model (LLM) based Multi-Agent Systems (MAS) utilizing explicit inter-agent trust modeling. When the internal trust coefficient ($\tau$) between a Custodian-Agent (holding sensitive data) and a Seeker-Agent is set to a high level (e.g., $\tau=0.9$) to optimize coordination efficiency, the Custodian-Agent systematically fails to enforce Minimum Necessary Information (MNI) principles. This…

The Trust Paradox in LLM-Based Multi-Agent Systems: When Collaboration Becomes a Security Vulnerability
Evaluated models: Llama 3 8B

PolyJailbreak Cross-Modal Safety Asymmetry

Published Analyzed Source: arXiv

The paper describes a reproducible black-box evaluation and attack framework, PolyJailbreak, for multimodal LLMs. It reports that uneven text-versus-vision safety alignment allows jointly optimized text and image inputs to bypass refusal behavior without model internals. The authors attribute this to visual alignment weakening textual refusal representations and to cross-modal fusion making harmful intent harder to separate from benign intent. These are paper-reported findings, not…

Multimodal Safety Is Asymmetric: Cross-Modal Exploits Unlock Black-Box MLLMs Jailbreaks
Evaluated models: LLaVA 1.5 7B, LLaVA 1.6 7B, Qwen-2.5-VL (7B)+5 more
  • LLaVA 1.5 7B
  • LLaVA 1.6 7B
  • Qwen-2.5-VL (7B)
  • Llama 3.2 11B Vision
  • GPT-4o
  • GPT-4.1
  • Gemini 2.5 Flash
  • Claude 3.7 Sonnet

Black-Box Confidence Exploit

Published Analyzed Source: arXiv

A vulnerability exists in the self-reflection and introspection capabilities of Large Language Models (LLMs) and Vision-LLMs that allows attackers to perform black-box adversarial optimization using only textual model responses. This technique, termed "Asking for Directions" (AfD), bypasses the need for access to gradients, logits, or continuous confidence scores. The attacker employs a hill-climbing optimization strategy where they present the target model with two candidate inputs (an…

Black-box Optimization of LLM Outputs by Asking for Directions
Evaluated models: Qwen 2.5 VL 3B Instruct, Qwen 2.5 VL 7B Instruct, Qwen 2.5 VL 72B Instruct+8 more
  • Qwen 2.5 VL 3B Instruct
  • Qwen 2.5 VL 7B Instruct
  • Qwen 2.5 VL 72B Instruct
  • Llama 3.2 11B Vision
  • Llama 3.2 90B Vision
  • Llama 3.1 70B Instruct
  • GPT-4o
  • GPT-4o Mini
  • GPT-5 Mini
  • Claude 3.5 Haiku
  • Claude 3.7 Sonnet

Leaked Bits Collapse Attack Queries

Published Analyzed Source: arXiv

Large Language Models (LLMs), specifically variants of GPT-4o, DeepSeek-R1, OLMo-2, and Llama-4, are vulnerable to accelerated adaptive adversarial attacks due to excessive information leakage in observable output signals. When these models expose "thinking processes" (Chain-of-Thought traces) or token-level log-probabilities (logits) to the end user, they leak significant mutual information $I(Z;T)$ regarding the model's safety state or hidden instructions. This leakage allows adaptive attack…

Bits Leaked per Query: Information-Theoretic Bounds on Adversarial Attacks against LLMs
Evaluated models: DeepSeek R1, GPT-4o Mini 2024-07-18, Llama 4 Maverick 17B+4 more
  • DeepSeek R1
  • GPT-4o Mini 2024-07-18
  • Llama 4 Maverick 17B
  • Llama 4 Scout 17B
  • OLMo 2 7B-1124
  • OLMo 2 13B-1124
  • OLMo 2 32B-0325

Agent Harassment Escalation

Published Analyzed Source: arXiv

Large Language Model (LLM) agents powered by LLaMA-3.1-8B-Instruct and Gemini-2.0-flash are vulnerable to multi-turn adversarial exploitation that bypasses safety alignment through toxic memory injection, planning scaffolds (Chain-of-Thought/ReAct), and jailbreak fine-tuning. Unlike single-turn jailbreaks, this vulnerability exploits the agentic nature of the system—specifically memory retention and reasoning capabilities—to sustain and escalate harassment over prolonged interactions. When…

Echoes of Human Malice in Agents: Benchmarking LLMs for Multi-Turn Online Harassment Attacks
Evaluated models: Llama 3.1 8B Instruct, Gemini 2.0 Flash 001

Graph-LLM Semantic Attack

Published Analyzed Source: arXiv

Graph-LLMs (Graph Neural Networks integrated with Large Language Models) utilized for representation learning on Text-Attributed Graphs (TAGs) are vulnerable to the Interpretable Multi-Dimensional Graph Attack (IMDGA). This vulnerability exists due to the non-decoupled nature of text encoding and graph message passing mechanisms. A black-box attacker can manipulate node classification predictions by executing a three-stage attack: (1) utilizing a word-level Topological SHAP module to identify…

Unveiling the Vulnerability of Graph-LLMs: An Interpretable Multi-Dimensional Adversarial Attack on TAGs
Evaluated models: Not reported

Personalized Disinformation Jailbreak Escalation

Published Analyzed Source: arXiv

Appending simple demographic persona details to prompts requesting policy-violating content can bypass the safety mechanisms of Large Language Models. This technique, referred to as persona-targeted prompting, adds details such as country, generation, and political orientation to a request for a harmful narrative (e.g., disinformation). This systematically increases the jailbreak rate across most tested models and languages, in some cases by over 10 percentage points, enabling the generation…

A Multilingual, Large-Scale Study of the Interplay between LLM Safeguards, Personalisation, and Disinformation
Evaluated models: Claude 3.5 Sonnet, Gemma 2 9B IT, GPT-4o+5 more
  • Claude 3.5 Sonnet
  • Gemma 2 9B IT
  • GPT-4o
  • Grok 2
  • Llama 3 8B Instruct
  • Mistral Nemo Instruct
  • Qwen 2.5 7B Instruct
  • Vicuna 7B v1.5

Refusal-Aware Embedding Jailbreak

Published Analyzed Source: arXiv

Large Language Models (LLMs), specifically open-weights models such as Llama-2, Mistral, and Vicuna, are vulnerable to a white-box adversarial attack framework termed RAID (Refusal-Aware and Integrated Decoding). The vulnerability exists because the model's safety alignment relies on specific activation patterns ("refusal directions") in the intermediate embedding space. Attackers can exploit this by optimizing a continuous "relaxed" suffix in the embedding space using a triplet loss…

RAID: Refusal-Aware and Integrated Decoding for Jailbreaking LLMs
Evaluated models: Llama 2 7B, Mistral 7B, Vicuna 7B

Research methodology

Entries summarize publicly available primary-source security research. Model names reflect only systems explicitly evaluated by the cited paper, and measurements are research-reported unless independent verification is stated.

Discovery and publication are separate: new papers enter a candidate queue, then require human review and explicit approval before publication. Updates follow that review process, not a guaranteed daily publishing schedule. “Last analyzed” is the newest analysis date among published entries; it does not mean that all recent research has been covered.